Vertical AI for regulated healthcare, built on kOS
Running inside regulated healthcare today. Claims and document workflows, a PHI-grade privacy posture, and decisions that carry an author, on the same EU-sovereign platform every KlusAI use case lands on.
The vertical product sits on top of the platform. Identity, tenancy, audit trails and data residency are already underneath it.
Built for work that is regulated before it is automated
Every capability carries the obligation it has to satisfy. In healthcare the obligation usually arrives first.
Document and claims workflows
Intake, extraction and review across referrals, claims and clinical correspondence, with every field traceable to the page it came from.
A reviewer checks the machine rather than trusting it, on the record.
PHI-grade privacy posture
Per-organization data residency, and a private mode that keeps inference on EU-resident compute instead of a cloud provider.
Where the data lives and where it is processed are both settings you can show.
Clinical de-identification
kp-deid removes personal data before text reaches a model, measured on the clinical domain rather than assumed from the legal one.
GDPR data minimisation applied at the boundary, not promised in a policy.
Decisions that carry an author
Every automated step records the actor, the model version and the reviewer, and the steps that need a human stop for one.
A tamper-evident trail is what turns an AI decision into an auditable one.
Scoped as a use case, not as a project
Two from the library, written up in full: what the outcome is, which obligations shape it, and how it is sized.
The de-identifier is ours, and it is measured on clinical text
Clinical text is its own problem: identifiers hide in dictated notes and free-text fields, and a redaction that leaks one national ID has failed. kp-deid covers the clinical domain in EuroPriv-Bench, our open benchmark, across 8 EU languages.
See the researchNational-ID leakage, lower is better
EuroPriv-Bench, verified July 26, 2026.
What the AI may touch is a setting, not a convention
Every tool the assistant can reach is declared: what it reads, what it writes, and which actions stop for a human to confirm. An organization handling patient data can show its regulator the configuration rather than describe the intent.
In production, not in decks
A top-3 Dutch health insurer is running AI-native delivery inside regulated healthcare.
Use cases scoped, built and running in production, under the privacy and audit obligations the sector actually carries.
Anonymized by agreement. We name the shape of the work, never the client.
Vertical AI products, built with domain partners
We build AI-native products for regulated European industries together with partners who own the domain expertise. Healthcare is the second vertical, and the model is the same one legal and notarial runs on.
The partner brings
- Domain expertise
- Distribution and customer relationships
- The regulatory reality of the practice
KlusAI brings
- kOS, in production and audited
- Models, including kp-deid
- Delivery, at a fixed fee per use case
Both hold
- A shared product
- Shared upside
- An IP model agreed before work starts
Own a healthcare domain?
Bring the domain. We bring the platform, the models and the delivery.
Message received!
We'll send you a confirmation email shortly.
Want to track your inquiry and access exclusive content?
Create your KlusAI accountStay in the loop
Weekly insights on production AI. No hype, just what works.
Follow us on LinkedIn
Stay updated on our latest AI work and insights.